Enterprise
Banking and insurance
In this sector an automation you can’t explain doesn’t survive internal review, whatever it saves. Everything I build here leaves a record that reads in review.
The sector’s needs
- A risk and reporting platform that aggregates positions across the core systems and regenerates any figure on demand
- Client data kept under one governed jurisdiction, with every operator, transfer and support path mapped and defensible
- Document pipelines that absorb the intake waves, onboarding, claims, credit files, with a decision trail instead of seasonal headcount
- Legacy core systems opened through governed interfaces, so data products ship without touching the platform that carries the contracts
- Models in production with drift measured, versions traced and every automated decision explainable to internal review
Delivered across Switzerland, remote-first, with on-site days by arrangement.
Scope boundary
The scope produces architecture, controls and factual evidence. It isn’t regulatory assurance, a prudential audit or legal advice: those conclusions stay with the right functions and qualified specialists.
Examples of what gets built
Client onboarding and KYC files
Identity papers, registers and forms extracted and cross-checked at intake volume, with confidence thresholds and a decision trail a reviewer can replay.
Document automation →An assistant grounded in the normative corpus
Directives, credit policies, product terms and circulars queryable in plain language, with the exact passage cited and directory permissions applied to every answer.
Secure AI assistants →Monitoring of transaction flows
Deviation models over transaction and declaration flows, factor contributions shown, thresholds governed and every alert traceable to its data.
Predictive analysis →Market-data platforms
Provider ingestion, documented transformations and datasets delivered on schedule, with freshness and completeness measured rather than assumed.
The data foundation →Control evidence
Processing register, access log and control evidence produced by the platform itself, with any figure regenerable on demand.
Search across contracts and endorsements
Finding an obligation by its meaning rather than its exact wording, across documents that share neither structure nor vocabulary.
Document search →Processing chains with a checkpoint
The repetitive steps of a case chained at portfolio volume, with a handler approving at the points where the decision commits and throughput measured.
Business agents →Core systems past end of support
The platforms carrying the policies outlive their support without a project that justifies replacing them. They move onto a version still supported, with the rollback written down.
Leaving end of support →Leaving proprietary virtualisation
The virtual estate carrying the management applications changes platform in waves, each step costed and documented for internal review.
Leaving proprietary virtualisation →Jurisdiction of client data
The record decides the option: operator, roles, contracts, locations, support access, subprocessors, transfers, retention, export, deletion and the exit from the former host.
Migration to hosting you chose →Provider reversibility
Ownership of access, exports readable without the original tool and transition documentation, prepared before internal review asks for them.
Reversibility pack →Access rights carried by roles
Every access to client or market data corresponds to a named role, with its justification and its latest review. The auditor reads the matrix; nobody rebuilds it for them.
Identity, directory and access →An infrastructure where every change keeps its author
Production and staging environments are described in reviewed code. A rule opened on a Friday evening shows up in review, and the history answers the compliance questionnaire without reconstruction.
Versioned infrastructure →Recovery objectives decided per service
Claims handling, pricing and accounting each receive a downtime and a data loss management has accepted. A timed restore and then a failover rehearsal prove the architecture holds them.
Backup and recovery →
Examples of solutions
The data foundation
Your sources reconciled in a governed model, with definitions, owners and controls.
Secure AI assistants
An assistant evaluated on your documents, with sources, permissions, processing and retention written down.
Document search
Finding the right page in the whole archive, in one question.
Document automation
Reading, sorting and extracting whatever arrives by email, PDF or scanner.
Business agents
Tasks chained end to end, with a person at the points that matter.
Predictive analysis
Forecasting demand, consumption or failure, from what you already measure.
Leaving end of support
Systems past their support date, moved before the incident rather than after.
Leaving proprietary virtualisation
Leaving a licensing model that changed under you, without losing the estate.
Migration to hosting you chose
Server, files or mail moved to an evidence-based choice, with rollback and exit defined.
Reversibility pack
Everything needed to leave a provider, or to carry on the day I stop.
Identity, directory and access
One account per person, rights carried by groups, technical secrets inventoried and revocations you can verify.
Backup and recovery
An isolated copy, a restore performed and timed, then recovery objectives decided per service and rehearsed.
Versioned infrastructure
Azure, Google Cloud or Proxmox infrastructure described in versioned files, with reviewed changes and a testable rebuild.
Relevant experience
Assignments I delivered, described without naming anyone. No logos and no testimonials: what I did, and with what.
Data-pipeline architecture on Google Cloud for an international index and financial-data provider
Airflow · dbt · BigQuery · Terraform
Systems and storage engineering at a global storage-solutions vendor
FC · iSCSI · NAS/SAN
AI extraction of data from financial announcements, in production at an international index provider
Vertex AI · BigQuery · prompts versionnés
Why me
- AI extraction running daily in production on financial announcements, not a conference demo
- Habit of environments where every transformation has to be documented and replayable
- Prompts kept and tested at every change, with the version used available for an internal review
- Swiss or on-premise hosting treated as an architecture decision, not a billing option
Neighbouring sectors
Energy and utilities
Forecasting, integration with the operating systems, and deviations caught in the measurements.
Healthcare and life sciences
Confidential assistants, research pipelines and document extraction, on data that doesn’t leave.
Retail and e-commerce
Demand forecasting, bounded pricing and assisted customer care, across multiple points of sale.
Test the fit: Banking and insurance
Describe the context, constraints and decision you need to make. The first conversation qualifies scope, boundaries and the next useful step.
Describe the situation